xlm-roberta-large-xnli-v3.0

This model is a fine-tuned version of joeddav/xlm-roberta-large-xnli on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.2434
• F1 Macro: 0.9286
• F1 Micro: 0.9287
• Accuracy Balanced: 0.9296
• Accuracy: 0.9287
• Precision Macro: 0.9288
• Recall Macro: 0.9296
• Precision Micro: 0.9287
• Recall Micro: 0.9287

Model description

More information needed

Intended uses & limitations

More information needed

Training and evaluation data

More information needed

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

• learning_rate: 9e-06
• train_batch_size: 8
• eval_batch_size: 64
• seed: 40
• gradient_accumulation_steps: 4
• total_train_batch_size: 32
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_ratio: 0.06
• num_epochs: 3

Training results

Training Loss	Epoch	Step	Validation Loss	F1 Macro	F1 Micro	Accuracy Balanced	Accuracy	Precision Macro	Recall Macro	Precision Micro	Recall Micro
0.2328	1.69	200	0.2811	0.8942	0.8942	0.8943	0.8942	0.8942	0.8943	0.8942	0.8942

eval result

Datasets	asadfgglie/nli-zh-tw-all/test	asadfgglie/BanBan_2024-10-17-facial_expressions-nli/test	eval_dataset	test_dataset
eval_loss	1.38	0.25	0.277	0.243
eval_f1_macro	0.583	0.925	0.91	0.929
eval_f1_micro	0.584	0.925	0.91	0.929
eval_accuracy_balanced	0.592	0.925	0.91	0.93
eval_accuracy	0.584	0.925	0.91	0.929
eval_precision_macro	0.595	0.925	0.91	0.929
eval_recall_macro	0.592	0.925	0.91	0.93
eval_precision_micro	0.584	0.925	0.91	0.929
eval_recall_micro	0.584	0.925	0.91	0.929
eval_runtime	50.84	0.65	0.123	0.51
eval_samples_per_second	167.193	1455.491	1531.96	1484.664
eval_steps_per_second	2.616	23.079	24.317	23.535
Size of dataset	8500	946	189	757

Framework versions

• Transformers 4.33.3
• Pytorch 2.5.1+cu121
• Datasets 2.14.7
• Tokenizers 0.13.3